Schemas and improvisation in Indian mu-sicRichard Widdess

1 IntroductionMany years ago the Indian classical singer Ritwik Sanyal was performing at amusic festival in India, in which he had planned to sing a song of his own com-position in the melodic mode or raga Bagesrı, and in a rare metrical structure(tala) of 9 beats. Warming up in the green room, while the previous artiston the programme was performing on stage, he found that his accompanist onthe drum pakhavaj was uncomfortable playing the 9-beat tala, so he instantlyre-composed the song in a more familar 12-beat metre. He then found thatthe previous singer was already singing Bagesrı. No audience likes to hear thesame raga twice in succession, so he re-composed the song again, in the sametala but a di!erent raga. He then gave a full 40-minute performance, in whichthe newly re-composed song was preceded and followed by extensive melodicand rhythmic improvisation. The performance as a whole can hardly havebeen planned in advance, still less rehearsed with the accompanist.

How are such feats of spontaneous composition and improvisation (on thepart of both musicians) possible? Nettl [1974] suggested that all music consistsof a sequence of fixed structural points. If these are close together, the music ispre-composed; but if they are further apart, the performer must navigate fromone to the next by interpolating blocks of material, whether spontaneouslycreated or selected from a memory bank. This model of improvisation (whichI have simplified for the sake of argument) fits Indian music well in many re-spects: the metrical cycle of tala provides fixed temporal points for action,especially the first beat of the cycle, called sam; a pre-composed block, or“compostion” (bandis), alternates with improvised episodes; the melodic gram-mar of the raga provides pathways from note to note and motivic material forfilling space within the tala cycle. But Nettl’s model (in this over-simplifiedform) seems a rather one-dimensional concept of performance, as if the per-former needs only to think about one thing at a time—now I’ll do this, nowthat—and needs only to make decisions about how to get to the next “struc-tural point” at the beginning of each “block”. Intuition would suggest that the

198 Richard Widdess

process of performance is more complex than this, with multiple parametersin play at any one time, and decisions being made at many moments along theway between structural points.

A number of studies have suggested that improvisation in Indian classicalmusic is less spontaneous than it appears to be, and is almost always relianton extensive memorised materials and procedures [Meer, 1980, Slawek, 1998,Napier, 2006, Zadeh, 2012]. But whether a singer truly improvises or not canbe hard to tell, and seems to vary between individuals; some, like Sanyal, areclearly adept at on-the-spot composition. Perhaps the only way to tell thateverything has not been worked out in advance, is where an improvised passagethreatens not to turn out as intended, and we can hear the performer adjustinghis materials in real time, in order to hit a structural point or to compensatefor not hitting it. It is the ongoing negotiation of conflicting requirements andstructures that makes performance of this kind particularly exciting for thelistener.

This paper focusses on a performance by the singer Ritwik Sanyal, accom-panied on the drum pakhavaj by Ashok Kumar Tagore, recorded in the SOASrecording studio in 1988.1 It represents the dhrupad genre of North Indianclassical music, which entails certain distinctive structural and stylistic char-acteristics. In this genre performance begins with a long, unscripted melodicimprovisation, the alap, sung with little metrical regularity, minimal percus-sion accompaniment, and non-lexical vocables in place of poetic text. This isfollowed by a pre-composed song, sung in the same mode or raga as the alap,set in a fixed metre (tala), with pakhavaj accompaniment and song lyrics inclassical Hindi. Although the song can be rendered in a very few minutes,its performance is normally extended to many times that length, through theinterpolation of improvised episodes between reprises of the song’s refrain. Itis this process of variation (laykarı) that I wish to examine here. My thesisis that both the singer’s improvisation, and the listener’s comprehension of it,depend on the simultaneous combination of pre-existing schemas, which enablethe singer to arouse, and the listener to feel, varying degrees of uncertainty,expectancy and resolution.

2 Interaction in Indian musicThis paper describes a performance in which a solo musician, with a singleaccompanist, addresses an unseen, future audience via a recording studio mi-crophone, rather in the manner of an orator pre-recording a broadcast speech.

1The performance is presented in complete transcription and analysed in Sanyal andWiddess [2004]. This paper re-considers aspects of the earlier analysis from a cognitiveperspective, with reference to schema theory. The drum pakhavaj is a double-conical drumwith two drum-heads, played with the hands.

Schemas and improvisation in Indian music 199

Leaving aside (for the moment) a degree of interaction between the two mu-sicians involved, the scenario seems to be that of a monologue rather than adialogue. I wish to argue, however, that in such a performance, the soundrecording represents only one, externalised side of an interaction in which theother side, that of the listener(s), is silent, internalised, and potential.

The performance of Indian classical music is often represented as a processof interaction, and the behaviour of both soloists and listeners in live concertsconfirms this perception. Informed listeners are active listeners, de-codingthe music as it unfolds, and they show this by physical and verbal responses.Listeners may respond to a particularly felicitous, technically impressive orbeautiful moment in the performance with a physical movement of hand orhead, and/or with a verbal response such as “Vah!”, “Sabas!”, or “Kya bat!”,all equivalent to “Bravo!” or “Wonderful!”. Clayton has observed [2007] thatlisteners’ physical responses often begin before a climactic moment has beenreached, showing that listeners anticipate the resolution of musical processes.

The communicative process is not, however, limited to an externalised per-formance on the part of the soloist, and a largely internalised response by theaudience. In addition there is an intermediary, in the form of the accompanist,playing a drum (tabla or pakhavaj ). There may also be a melodic accompanist,playing the bowed lute sarangı or a small hand-pumped harmonium, thoughneither of these instruments is used in the performance examined here. Bothrhythmic and melodic accompanists provide a kind of running commentaryon the soloist’s performance, which we could understand as an externalisedrealisation of an informed listener’s real-time internal modelling of the perfor-mance. Thus the soloist does not interact with the accompanist alone, butwith the expectations of all listeners including the accompanist.

Three types of musical expectation have been distinguished [Huron, 2006](pp.231, 235) [Widdess, 2011]: veridical expectations, where the listener hasheard the same music before and knows precisely what comes next; schematicexpectations, where the listener can predict what is likely to come next on thebasis of familiar melodic, rhythmic or formal patterns; and dynamic expecta-tions, expectations based on the perception of patterns emerging as the musicunfolds. These categories undoubtedly overlap, and all depend on repetitionand memory. In the case of North Indian classical music, where “improvisa-tion” is emphasised more than the repetition of pre-composed pieces, schematicand dynamic expectations may be more significant than veridical.2 In partic-ular, I suggest that expectations are generated by cognitive schemas, whether

2But see Napier’s discussion of the use of inherited material in performance [2006]. Mycomments refer exclusively to North Indian classical music; veridical expectations may bemore significant in the case of South Indian classical music, where composed items are moreimportant than in the North.

200 Richard Widdess

these are learned through cumulative listening or generated dynamically withina single performance.

3 Schemas in musicA schema has been defined as “[a mental structure] formed on the basis of pastexperience with objects, scenes or events and consisting of a set of (usually un-conscious) expectations about what things look like and/or the order in whichthey occur” [Mandler, 1984]. A schema consists of a number of variables orcategories in a certain relationship. For example, in English the verb to writeevokes a schema in which a person guides an implement aross a surface soas to leave a trace, usually representing language [d’Andrade, 1995] (p.123).Person, implement, surface, trace and language are all variable categories, andthe schema articulates a spatial and temporal relationship between these cat-egories. Schemas3 may be culturally nuanced: in some cultures the “writing”schema includes drawing pictures, in others it refers only to representations oflanguage (ibid.).

Music seems to be composed of a large variety of specialised schemas, someof which may be explicitly formalised as music theory, and others not [Sny-der, 2000] (p95). They consist of categories including pitches, beats, rhythmicdurations, contours, themes, chords, stylistic elements etc., between whichthey establish static or temporal relationships: pitch intervals, scales, key re-lationships, melodic contours, modes, metrical or formal structures, styles andgenres etc. They can be hierarchically combined, and generate expectations.Cognitive schemas are not immutable, and can be modified to accommodatenew situations and experiences. Their existence in listeners’ perception of mu-sic can be empirically verified, as Rosner and Meyer showed in a pioneeringstudy [1982]; and their significance in Western art music has been discussed[Gjerdingen, 1988, Gjerdingen, 2007, Byros, 2009, Byros, 2012]. For schemasin Indian music, see Widdess [2011]; Zadeh [2012].

Schemas in music include both relatively fixed patterns (“scripts”, “formu-lae” or “ri!s”) and highly flexible structures (“plans”, “templates”) [Schankand Abelson, 1977]. The former tend to be relatively small-scale and to be em-bedded within the latter, but there is no definitive boundary between the twotypes. Schemas of both types can be combined in innumerable ways duringcomposition and improvisation, thereby constructing extended and complexpieces or performances. Thus a jazz performance, for example, might combinea harmonic schema (chord sequence) with a melodic script (the song tune), atemporal schema (metre), pitch schemas (key, scale, mode), a verbal script(lyrics), a formal template (song—solos—song), and a stylistic schema in-

3In Greek the plural of schema is schemata, but since this is not always recognized byEnglish speakers as a plural, I use the English plural form schemas.

Schemas and improvisation in Indian music 201

cluding aspects of sound, playing/singing techniques, ornamentation, rhythm,melodic conventions etc. that are typical of jazz. To consider these as separateschemas is no doubt an analytical approach: to the performer and listener, atthe moment of performance, they appear as a single blended entity.

I want to suggest that performance of Indian classical music can involvecombining schemas of di!erent kinds at the moment of performance, in a rel-atively unpremeditated way, and that this is what we mean when we callsuch music “improvised”. These schemas include melodic, text and rhythmicschemas, and both fixed scripts and overarching plans. I will suggest that thiscombination of schemas is possible because some are su"ciently flexible to beadapted to others that are less flexible. It is not that the combination of Aand B has to be mapped out in advance to make sure that they fit; rather thatB can be modified to fit A, or vice versa, as the combination unfolds. It is alsopossible to overlap schemas so that their boundaries do not coincide; but suchoverlap may create the expectation of subsequent reconciliation. This processof schema combination enables communication and interaction between soloist,accompanist(s), and silent but active listeners.

4 The schematic basis of performanceThe performance discussed here is based on a pre-composed song, in the raga(mode) Multanı, and in the tala (metre) Cautal. Figure 1 shows the first phraseof the song, occupying the first metrical cycle, which can be repeated; most ofthe ensuing improvisation is based on this first phrase. It establishes a numberof schemas that remain in play throughout the performance: principally (1)text, (2) tala and (3) raga.

(1) The text, Tuhi bidhata lokapati [namo namo], “I worship thee, theOrdainer and Lord of the Universe”, is a script that can be varied onlyminimally in improvisation, by the selection and repetition of groupsof words. As far as possible in the heat of improvisation, words arekept intact and not reduced to separate syllables. Words and groupsof words can be repeated, however.

(2) The metrical schema (tala) is a cycle of 12 beats, of which the first,called sam, is the most emphasised (it is in e!ect both the first andlast beat of the cycle). The rotation of the cycle is indicated inperformance with hand gestures—claps and waves—according to aconventional pattern that divides the cycle into 2-beat segments(shown above the sta! in figure 1). These gestures assist the drumaccompanist, the audience, and the singer himself, in keeping count oftime. They form an embodied metrical schema, a template withinwhich an infinite variety of rhythms can be created.

202 Richard Widdess

(3) The melodic mode or raga determines

a. a scale (here a non-diatonic heptatonic series with !2, !3, "4 and!6);

b. a pitch hierarchy: here 1 and 5 are strong, !2 and !6 weak; and

c. ascending and descending pitch sequences: here !2 and !6 appearin descent only – and in this phrase, not at all.

Figure 1: First phrase of composition in raga Multanı, as sung by R. Sanyal

After singing the composition, the singer improvises on it, periodically re-turning to the beginning of the composition. In this style of improvisation,called laykarı, the main schemas remain in play: raga, tala, and text. Whatchanges is the relationship in time between these schemas. The tala and temporemain constant, but there is an increase in rhythmic density, that is, the num-ber of notes and syllables per beat, so that both the melody and the text passby more quickly in the improvisation than in the original composition. Eachperiod of improvisation must end after one or more cycles, on beat 1, the sam.The drum accompanist improvises simultaneously with the singer, and mustalso conclude on beat 1 (sam), accurately predicting the end of each improvi-sation. Figure 2 shows an example of laykarı improvisation by Sanyal, basedon the first phrase of the composition shown in figure 1.

Is laykarı improvised, in the sense of being composed at the moment of per-formance? Some singers reduce the risk of error by calculating and memorisingit in advance; others say that improvised laykarı sounds more spontaneous ande!ective, and it is therefore worth the risk of missing the sam. Ritwik Sanyalbelongs to the latter viewpoint. When I asked him how he improvises laykarı,

Schemas and improvisation in Indian music 203

Figure 2: Example of Sanyal’s laykarı improvisation on the Multanı composi-tion

however, he could not specify any particular technique: he said that one learnsit by imitating one’s teacher, and “when I am in the right mood, it works”.Analysing examples like this one, however, enables us to suggest some of thecompositional techniques involved and to identify the important role of schemacombination and re-combination.

5 TihaısThe most critical part of an episode of improvisation is its ending, since thismust be successfully timed to end with a return to the beginning of the com-position and an emphatic arrival on sam. Before that point, the singer seemsto have almost limitless freedom; but the challenge is to time the end of the“free” part of the improvisation in such a way as to arrive e!ectively at sam.

In most of Sanyal’s improvisations, the last few beats of each episode areconstructed in the form of a tihaı. This common feature of Indian music, prob-ably originating in dance, comprises a rhythmic motif repeated three times,ending on or just before sam. The motif for a tihaı can be of any length,but it must begin at the correct point in the metrical cycle in order to end

204 Richard Widdess

at sam after three repetitions, with or without a rest between repetitions. Atthe same time the rhythmic motif must be combined with a suitable melodicmotif, derived from the raga, and with a segment of the song-text; indeed therhythm of the text segment being sung at the time may suggest to the singerthe rhythm of a tihaı. Tihaıs, I suggest, are particularly revealing about theprocesses of real-time schema combination in improvisation.

Figure 3 shows a sample of tihaı patterns used in this performance, lastingin total 1, 1.5 and 3 beats (longer ones, up to 5 beats, are also used in thisperformance: see Sanyal and Widdess [2004] (p263)). In each case, the singerneeds to sing the first syllable of the song, namely Tu, on the concluding sam,in order to return to the first phrase of the composition (see figure 1). Thesetihaıs therefore all end just before sam. In the 3-beat tihaıs, the rhythmicmotif occupies 1 beat, repeated 3 times (T1, T2 and T3 in figure 3). Thisrhythmic motif is used on three occasions in this performance, but each timeit is combined with di!erent chunks of text, and the melodic content is alsodi!erent each time. The text-chunk has to be a valid word or group of wordsfrom the song-text, and the melody has to conform to the structure of the raga.The three-beat tihaıs shown here exemplify a rhythmic script—a small-scale,relatively fixed formula—conforming to a template common to all tihaıs. Butthe rhythm, melody and text of each realisation of this three-beat tihaı belongto independent schemas that are combined at the moment of performance.

A tihaı can be spontaneously adjusted in performance. For example, if toofew syllables remain in the current “chunk” of text, some may have to berepeated (e.g. the first syllable of u-ta-ra in cycle 31); or if the singer mis-times the start of the tihaı, it can be discreetly stretched or compressed tofit the time available. The singer can even subvert the schematically-induceddynamic expectations of the listener (and drummer) by deliberately arrivingearly or late for the sam (this is called atıt or anaghat respectively).

6 Laykarı improvisationThe example of laykarı improvisation shown in figure 2 occupies two completemetrical cycles and culminates on sam of the third cycle; each line of notationhere represents one cycle. In each cycle, the improvisation is structured inthree phases, numbered 1–3 in figure 2, of which the third is occupied bya tihaı. This three-fold division of the cycle is something Sanyal often doesin longer tals, though he was unaware of it until I asked him about it; hetends to take a di!erent text chunk in each phase, and here he also situatesbreaths between them. But the phase divisions do not correspond to equaldivisions of the cycle: the three-phase schema is flexibly mapped onto themetrical schema, showing that they are independent schemas, and that thethree-phase schema can override the inflexible structural boundaries inherent

Schemas and improvisation in Indian music 205

Figure 3: Sample tihaı structures in Sanyal’s Multanı performance

in the metrical schema.The improvisation in the first phase of cycle 1 comprises two short exclam-

atory phrases, separated by rests, as if sounding a warning to the drummer(and listener) that something complex is about to happen. The first excla-mation begins in the last beat of the preceding cycle, overlapping the cycleboundary. The second phase, at beat 5, introduces an iambic rhythm whichtransgresses the boundary, marked by the dotted barline, between two-beatsegments at beat 7. This beat is denoted by a silent wave in the embodiedmetrical schema, and is an important structural point in the tala cycle, thebeginning of the second half of the cycle. Sanyal normally marks this beatwith a new syllable, but here he allows the iambic rhythm to flow across it.The third phase has a similar boundary violation at the 4th clap beat (beat10), and its text chunk implies a continuation into the next cycle.

We need only examine the tihaı at the end of the first cycle here to see thecomplexity of the challenges that the improvising singer faces in combiningthe di!erent schematic elements in play. The tihaı motif is based on the word

206 Richard Widdess

lokapati, but when we arrive at the end of the cycle we are only half-waythrough this word (lokapati lokapati loka - ?). If the singer wished to “come tosam” at beat 1 of the next cycle, he would have to leave the word unfinishedand sing Tu. More elegant is Sanyal’s spur-of-the-moment solution: he startsa new phase of melodic and rhythmic improvisation on sam, and allows theword [loka]-pati to overflow into the new metrical cycle, displacing the expectedsyllable Tu to the third 16th (an example of anaghat : see above). Postponingthe expected conclusion on sam, he continues his improvisation for a secondcycle.

The disjuncture here is also reflected in the melody. The melodic scriptof the tihaı (F" A! G)is interrupted after the A! by the new improvisationbeginning on E!: A! is normally followed by G in this raga, and this is theonly occurrence of the interval A! – E! in the entire performance (it wouldnormally be considered a grammatical error). The expected resolution on G ispostponed to the third 16th, at the syllable Tu. At this point both text- andraga-schemas are resolved at the same time, but not in synchrony with themetrical schema; and the G is as much part of the new phase of improvisation,beginning two 16ths earlier, as it is the conclusion of the tihaı. Thus melodicand textual schemas are allowed to overlap each other and the rigid boundaryof the metrical schema.

This tension between melodic, textual and metrical schemas remains to beresolved at the end of the second cycle, and is e!ected by a second tihaı. Thistime the tihaı occupies five beats, and is based on the three-syllable wordbidhata; it uses the same melodic script—F" A! G—as the previous tihaı,leading to the first note of the song melody, F". But here the three iterationsof the tihaı formula are not identical: T3 is longer, and is also melodicallyand rhythmically di!erent from T1 and T2. Here we have an example of thespontaneous adjustment of a tihaı to ensure that the singer arrives correctlyon the sam, on the correct beat, the correct syllable (Tu) and the correct note(F" – the first note of the song). Here all the schemas in play are brought intoalignment once more.

Figure 4 analyses the progressive unfolding of the tihaı. On line 1 it showsthat if the tihaı were continued in the most straightforward way, it wouldend correctly on sam, but on the wrong syllable (-ta not Tu) and on thewrong pitch (G not F") (the cross-head notation in figure 4 shows the inferredcontinuation). In order for the syllable Tu to be sung on the sam beat, thetihaı needs to end just before sam if it is not to be truncated. Accordingly,Sanyal brings in T2 half a beat earlier than might have been expected. Butnow the syllable Tu is set to fall half a beat too soon, if the pattern of the tihaıso far is continued to the end. T3 is therefore stretched to fill the remainingtwo beats of the cycle (figure 4, line 3). In T3 the rhythmic pattern changes to

Schemas and improvisation in Indian music 207

fill the space (its three successive notes becoming progressively shorter), andthe melody also changes so as not to prolong the weak pitch A!, which wouldbe a grammatical mistake in this raga. At the same time the pakhavaj playerplays a loud tihaı of his own as if to keep singer on track; but this tihaı startson beat 9, so it is di!erent in length and structure from the singer’s tihaı,coinciding with it only at the end point. (The accompanist has foreseen theconclusion on sam, but not the structure of the tihaı; rather he has assumedthat the tihaı would begin on beat 9, as it did in the previous cycle.)

Figure 4: Spontaneous modification of tihaı in Sanyal’s Multanı performance

The singer successfully adjusts his tihaı as it unfolds to conclude on thecorrect beat, syllable and pitch, simultaneously with the accompanist, andin fulfilment of our metrical and other expectations. At this point, the fixedboundary of the metrical schema overrides all other schemas, which are ad-justed to bring them into synchrony with it. It is at moments like this thatexpert members of a live audience would be likely to say “Vah! Sabas!”, orindicate by silent movements their anticipation and appreciation of the musi-cians’ successful arrival at sam, and their resolution of the tensions caused byallowing schemas to slip out of synchrony earlier in the improvisation.

7 ConclusionsI argue that, in Ritwik Sanyal’s laykarı, the physical tala-gestures (shownat the top of each example) embody a schematic plan of the metrical cyclethat maps and structures everything that happens within it. What happensis that schemas controlling melody, rhythm and text, are combined not only

208 Richard Widdess

sequentially, as ready-made chunks or blocks, but also simultaneously. Thisrequires their mutual adjustment in real time as the combination unfolds, moreflexible schemas (including melodic contours, rhythmic motifs, tihaıs and textchunks) being adjusted to less flexible ones (principally the tala cycle itself).

This model of improvisation would help to account for the kind of unpre-mediated performance that I described at the start of this chapter. But weneed not assume that those performers who rely more on memorized materialare any less reliant on schemas. As Treitler (following Bartlett) argued [1974],memory recall in music is a reconstructive or generative process, and Rubin[1995] has shown that “combined constraints”, including rhythm and melody,are among the means by which memorised text is recalled or reconstructedin oral performance. Schemas are equally important for listeners to form anunderstanding of what is heard, whether by recognizing schemas that theyalready know, or by perceiving dynamically the schematic structure of musicas it unfolds, and thereby participating in the interactive process of musicalcommunication. I suggest that empirical and analytical study of the schematicstructure of orally composed and transmitted music, and of processes of schemacombination, would illuminate our understanding of phenomena that we im-precisely characterise as improvisation.

BIBLIOGRAPHY[Byros, 2009] Vasili Byros. Towards an “archaeology” of hearing: Schemata and eighteenth-

century consciousness. Musica Humana, 1(2):235–306, 2009.[Byros, 2012] Vasili Byros. Meyer’s anvil: Revisiting the schema concept. Music Analysis,

31(3):273–346, 2012.[Clayton, 2007] Martin Clayton. Time, gesture and attention in a“ khyal” performance.

Asian Music, 38(2):71–96, 2007.[d’Andrade, 1995] Roy G. d’Andrade. The development of cognitive anthropology. Cam-

bridge University Press, 1995.[Gjerdingen, 1988] Robert O. Gjerdingen. A classic turn of phrase: Music and the psy-

chology of convention. University of Pennsylvania Press Philadelphia, 1988.[Gjerdingen, 2007] Robert O. Gjerdingen. Music in the galant style. Oxford University

Press, 2007.[Huron, 2006] David Brian Huron. Sweet anticipation: Music and the psychology of expec-

tation. The MIT Press, 2006.[Mandler, 1984] Jean Matter Mandler. Stories, Scripts, and Scenes: Aspects of Schema

Theory. ERIC, 1984.[Meer, 1980] Wim van der Meer. Hindustani music in the 20th century. Martinus Nijho!,

The Hague, 1980.[Napier, 2006] John James Napier. A subtle novelty: Repetition, transmission and the

valorisation of innovation within North Indian classical music. Critical Studies in Impro-visation/Etudes critiques en improvisation, 1(3), 2006.

[Nettl, 1974] Bruno Nettl. Thoughts on improvisation: A comparative approach. TheMusical Quarterly, 60(1):1–19, 1974.

[Rosner and Meyer, 1982] Burton S. Rosner and Leonard B. Meyer. Melodic processes andthe perception of music. In D Deutsch, editor, The psychology of music, pages 317–341.Academic Press, New York, 1982.

Schemas and improvisation in Indian music 209

[Rubin, 1995] David C. Rubin. Memory in Oral Traditions: The Cognitive Psychology ofEpic, Ballads, and Counting-out Rhymes. Oxford University Press, 1995.

[Sanyal and Widdess, 2004] Ritwik Sanyal and Richard Widdess. Dhrupad: Tradition andperformance in Indian music. Ashgate Publishing, Ltd., 2004.

[Schank and Abelson, 1977] Roger C. Schank and Robert P. Abelson. Scripts, plans, goals,and understanding: An inquiry into human knowledge structures. The Artificial Intelli-gence Series. Psychology Press, 1977.

[Slawek, 1998] S. Slawek. Keeping it going: terms, practices and processes of improvisationin Hindustani instrumental music. In Bruno Nettl and Melinda Russell, editors, In thecourse of performance: Studies in the world of musical improvisation. University ofChicago Press, 1998.

[Snyder, 2000] Bob Snyder. Music and memory: an introduction. The MIT Press, 2000.[Treitler, 1974] Leo Treitler. Homer and Gregory: The transmission of epic poetry and

plainchant. The Musical Quarterly, 60(3):333–372, 1974.[Widdess, 2011] Richard Widdess. Dynamics of melodic discourse in Indian music: Budha-

ditya Mukherjee’s alap in rag Puriya-Kalyan. In Michael Tenzer and John Roeder, editors,Analytical and cross-cultural studies in world music, pages 187–224. Oxford UniversityPress, 2011.

[Zadeh, 2012] Chloe Zadeh. Formulas and the building blocks of thumrı style–a study in“improvised” music. Analytical approaches to world music, 2, 2012.

Richard WiddessSchool of Oriental and African Studies, University of London, UK.rw4@soas.ac.uk